Overview
This research investigates the identification of operating regimes in a 10-MW Kaplan hydrogenerator, specifically the Porjus U9 unit. The methodology centers on fusing synchronized measurements of distributed air-gap flux and rotor current. The goal is to develop descriptors that encompass both electrical loading and electromagnetic field behavior for reliable monitoring of hydroelectric generators.
Research Context
Reliable monitoring of hydroelectric generators requires comprehensive descriptors capable of capturing both electrical loading and the behavior of the electromagnetic field within the machine. This study specifically addresses this need in the context of a Kaplan hydrogenerator. The Porjus U9 10-MW Kaplan hydrogenerator served as the subject for this investigation, providing a specific operational environment for data collection and analysis.
Approach
The study utilized synchronized measurements from two distinct sensor types: ten stator-mounted Hall probes and six rotor-current channels. These measurements were collected from the Porjus U9 10-MW Kaplan hydrogenerator. Seven steady guide-vane-opening settings were considered as distinct operating points for analysis.
Each recorded data segment, lasting 300 seconds, was subsequently divided into 1-second windows for granular analysis. Three categories of features were extracted from these windows:
- Spatial Fourier descriptors: Derived from the circumferential air-gap field measurements.
- Probe-wise temporal flux indicators: Extracted from individual Hall probe readings over time.
- Channel-wise RMS rotor-current features: Calculated from the six rotor-current channels.
To understand the relationship between these feature groups and the operating point, correlation analysis and principal component analysis (PCA) were employed. For the supervised identification of the guide-vane-opening state, three classification models were evaluated:
- Random Forest
- Radial-basis-function support vector classification (SVC-RBF)
- Multilayer perceptron
Findings
- RMS rotor-current features primarily tracked the loading axis of the hydrogenerator.
- Magnetic-flux features revealed complementary information, specifically related to spatial imbalance, waveform distortion, and weak low-frequency modulation.
- Spatial descriptors alone provided limited separability for operating regime identification, resulting in test accuracies below 27%.
- Rotor-current features alone achieved test accuracies of approximately 84-85%.
- Combining both flux and current information yielded the most discriminative representation.
- The SVC-RBF model, when utilizing combined flux and current information, achieved a 99.5% test accuracy and macro-F1 score for identifying guide-vane-opening states.
Why This Matters
The results indicate that distributed air-gap magnetic sensing, when integrated with rotor-current measurements, can support accurate and interpretable data-driven monitoring of Kaplan hydrogenerator operating regimes. This combined approach offers enhanced capabilities for distinguishing between different operational states.